scyld-cluster-conf -- load or save the cluster configuration file.



[-h] [-v] [-q] [[-c | --config] CONFIG] [--base-url URL] [[-u | --user] USER[:PASSWD]] {load, save} ...


-h, --help

Print usage message and exit. Ignore trailing args, parse and ignore preceding args.

-v, --verbose

Increase verbosity.

-q, --quiet

Decrease verbosity.

-c, --config CONFIG

Specify a client configuration file CONFIG.


--base-url URL

Specify the base URL of the ClusterWare REST API.

-u, --user USER[:PASSWD]

Masquerade as user USER with optional colon-separated password PASSWD.



Load CLUSTER_CONFIG as the new configuration file, optionally loading only nodes.


Parse the file, but do not alter the database.


Ignore other settings and only load networks.


Ignore other settings and only load nodes.


Save the current configuration file to file CLUSTER_CONFIG.


The scyld-cluster-conf command is primarily used to load a cluster configuration into ClusterWare including the PXE boot network definition(s) and the node definitions. A minimal useful configuration file consists of at least an iprange and one or more nodes:

node 08:00:27:A2:3F:C9

The first IP address in the iprange will be used to identify a local interface on the head node in order to find networking details such as the network mask. The DHCP range will be assumed to cover from the first IP up to the network broadcast address, but a "last" address can also be provided to limit that range:

nodes 10
node 08:00:27:A2:3F:C9
node 08:00:27:A2:E4:A2

Note that the node count can be provided in the file and a warning will be printed if more than that many nodes are defined in the file. The netmask can also be supplied as shown in the iprange line. Nodes will be numbered in order starting with index 0 but a line with no MAC address will act as a placeholder meaning this file would define nodes n0 and n2.


If multiple MAC addresses are included for a single node, only the first will be used.

Alternatively network definitions can specify where the node numbering actually starts:

node 08:00:27:A2:3F:C9
node 08:00:27:A2:E4:A2

This configuration file still defines a DHCP range of 100 IP addresses, now the nodes will be numbered starting with n1. In more complicated network configurations compute nodes may be split among multiple subnets:

node 08:00:27:A2:3F:C9
node 08:00:27:A2:E4:A2

21 via gw
node 08:00:27:FE:A3:22

The first network definition will be limited to 20 IP addresses based on the first index of the second network definition. For networks that are not locally accessible to the head node(s), such as in this case, the configuration file can also specify an optional route and compute node gateway. The route is specified through the via keyword and is only used to identify the appropriate interface for the DHCP server to listen to at run time. The gateway (gw) should be on the compute node network and will be provided to the booting nodes such that they can reach the head node cluster. A DHCP relay should be configured to forward DHCP traffic from the remote compute nodes to the head nodes and vice versa, and should populate the giaddr field of the DHCP request with an address on the compute node subnet. For directions on configuring DHCP relays, please see your switch or operating system documentation.

When defining multiple networks they must be defined in order of node indexing. Node indexes and IP addresses are assigned based on the most recently defined network so the above example defines 3 nodes, n1, n3, and n20. Additional nodes added dynamically will be assigned the lowest available index and the corresponding IP address.


Note that loading a cluster configuration will completely overwrite any existing configuration, including deleting all previously defined nodes.


We suggest restarting the clusterware service on all head nodes after loading a new cluster configuration.


scyld-cluster-conf save /root/cluster-conf-bak

Save a copy of the current network configuration and node list.

scyld-cluster-conf load /root/cluster-conf-new

Replace the existing node definitions with ones loaded from the /root/cluster-conf-new file.


Upon successful completion, scyld-cluster-conf returns 0. On failure, an error message is printed to stderr and scyld-cluster-conf returns 1.